ss-coding-workflow
Pass
Audited by Gen Agent Trust Hub on Jul 13, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill acts as a thin orchestrator, delegating all code modification and review tasks to specialized sub-skills (e.g., ss-coding, ss-create-branch, ss-create-pr), which limits its direct exposure to sensitive operations.
- [SAFE]: Explicitly mandates that the agent must 'Never relax safety guardrails' and specifies that automation flags must not bypass hard rules built into the underlying skills.
- [SAFE]: Implements a structured 'Review-Acceptance Loop' driven by review verdicts, ensuring that code changes are verified and that findings are addressed or explicitly justified before delivery.
- [SAFE]: Includes pre-flight checks to ensure that required developer tools (GitHub/GitLab CLIs) are available and that inputs are reachable before starting the workflow.
- [SAFE]: The skill ingests untrusted data from external plans or change instructions (SKILL.md). While this creates an indirect prompt injection surface, the orchestrator includes mitigating controls by delegating work to skills with their own guardrails and enforcing a mandatory review loop.
- Ingestion points: Inputs (plan path, change instruction, URL) in SKILL.md.
- Boundary markers: Not explicitly defined in this orchestrator.
- Capability inventory: Orchestrates code writing and PR creation via sub-skills.
- Sanitization: Delegated to underlying specialized skills.
Audit Metadata